In a simple network of a tree topology, multicast group information may be learned and forwarded through an Internet Group Management Protocol (IGMP) and an IGMP Proxy function. By deploying the IGMP Proxy function, the overhead of the protocol messages is reduced effectively, and the bandwidth is saved. However, a precondition of using the IGMP Proxy is that the network has a tree topology, in which a clear upstream and downstream relationship exists between devices.
Daisy chain networking is a semi-ring network formed by splicing pseudo wires (PWs) in a hierarchical virtual private LAN service (H-VPLS) mode. As shown in FIG. 1, through protocol independent multicast (PIM), a designated routing device is elected among the routing devices M and N on the top. For example, the routing device M is elected for forwarding multicast data. The switching devices A, B, and C below enable IGMP Snooping and IGMP Proxy to forward multicast data to terminal users as required. In a ring network, the enabling of the IGMP Proxy leads to the following problem: Switching devices A and C are upstream devices of the switching device B, and also downstream devices of the switching device B; after receiving a Join message destined for 225.1.1.1 from the user, the switching device B sends the message to the switching devices A and C; the switching devices A and C also send the Join message destined for 225.1.1.1 to their respective upstream device, including the switching device B; consequently, the switching device B believes that three of its ports require the data of 225.1.1.1, but in fact, only one user requires the data.
A solution to the foregoing problem in the conventional art is to select a querier as a unique upstream device, such as routing device M, by means of IGMP querier election. Other devices meet the following logical relationships strictly: The switching device A is the upstream device of the switching device B, the switching device B is the upstream device of the switching device C, and the switching device C is the upstream device of the switching device D. In this way, the correct topology relationship is ensured between devices, and the IGMP Proxy function can be deployed normally. However, the solution in the conventional art strictly restricts that only one upstream device exists, and, once the ring network is faulty, protocol convergence needs to be performed again to set up a correct upstream and downstream relationship. For example, when the switching device B is faulty, the multicast service does not return to normal until the relationship ensuring the switching device D as an upstream device of the switching device C is set up again, which sacrifices the advantage of the ring network characterized by fast recovery from a fault.